# US short-form trend research SOP

Use this process to turn current TikTok and Instagram Reels into a searchable library of creative patterns and original ideas. Add YouTube Shorts when it helps the brief. Run it on demand; no recurring collection is implied.

The reusable agent instructions are in `us-trend-research` skill. Detailed selection rules live in method reference, and verified API contracts and costs are in endpoint reference.

## Research run

1. **Set the brief.** Record the topic, intended audience, US relevance, platform, purpose, date window, and credit budget. Default to seven days and a maximum of 40 credits per manual run unless another authorized limit exists.
2. **Find candidates.** Combine recent feeds, topic searches, and a small creator watchlist. Use the supplied ScrapeCreators credential through the secret store or environment. Keep it out of reports and source code.
3. **Check the evidence.** Deduplicate posts; verify their publication dates; record available counters and when they were fetched. Keep older examples separate if widening the window to 30 days.
4. **Qualify US relevance.** Record whether the source verifies a US trend, supports US relevance, or merely came through a US proxy. A proxy setting does not prove a US audience.
5. **Choose useful patterns.** Compare similar platforms, topics, post ages, and durations. Look for independent creators repeating a mechanism. One large post remains a candidate until corroborated.
6. **Review the content.** Inspect the opening, story, sound, editing, payoff, and CTA. Use captions to support spoken details, and direct audiovisual review for visual and audio claims. Mark missing evidence clearly.
7. **Build the adaptation.** Write original hook options and a beat outline using the transferable mechanism, relevant proof, and a suitable CTA. Link to the source inspiration.
8. **Save and report.** Add the entry, preserve previous snapshots and human edits, and summarize the strongest patterns plus the limits of the scan.

## What a library entry contains

| Group | Content |
| --- | --- |
| Source | Creator, platform, canonical URL, publication date, capture date, query/source |
| Performance | Available views, likes, comments, shares and saves; metric provenance; prior snapshots when available |
| US evidence | Label, supporting source, period, and whether it describes the trend, creator, or audience |
| Hook | Spoken opening, opening visual, on-screen text, audio, intended audience promise |
| Structure | Curiosity gap, stakes, ordered beats, observed attention devices, payoff, CTA, loop |
| Explanation | Why the mechanism may work, alternative explanations, and confidence per claim |
| Adaptation | New hooks, original beat outline, required proof/assets, CTA, test hypothesis |

Keep “observed,” “inferred,” and “unknown” separate. A transcript-only record can be useful, but it cannot claim to have reviewed the visuals. A title or caption is not automatically the spoken hook.

## Interpreting trends

- **Candidate:** A promising post or feed result with limited corroboration.
- **Corroborated pattern:** At least three independent creators use the same identifiable mechanism within the selected window. This is a research rule, not a platform guarantee.
- **Growing:** Repeated live measurements show growth over a stated interval. Acceleration needs at least three comparable snapshots.
- **Saturated:** The inspected sample shows heavy repetition plus evidence that newer comparable examples are weakening or adoption has stopped expanding. Popularity by itself does not establish saturation.
- **Evergreen inspiration:** Useful creative outside the freshness window or without current momentum evidence.

Assess novelty relative to the inspected sample. Compare within platform, topic, age, and duration cohorts. Avoid combining raw views from different platforms into a universal ranking.

One snapshot proves counts at capture time. Views divided by post age is a rough lifetime average. Neither proves recent growth, retention, or why the content succeeded. Unknown metrics stay blank rather than becoming zero.

## Refresh and improve

For an authorized refresh, repeat a bounded discovery pass, enrich only new or changed candidates, and append live metric snapshots for the shortlist. Preserve dates and cache provenance. If few fresh results exist, explicitly widen to 30 days or return a smaller current set.

Use the endpoint guide's failure rules: no blind retries, no pagination without new results, and no automatic credit purchases or long-transcription upgrades. Save partial results when a source is unavailable.

After an adaptation is published through the normal authorized workflow, use account-owned analytics to compare hook variants and outcomes. Retention, completion, clicks, and conversions require appropriate analytics; public views and likes cannot substitute for them.

## From research to a finished script

This workflow adapts the topic research, writing, review, and packaging principles in the supplied AI Avatar Video Creation SOP. Avatar creation, voice cloning, vendor purchases, the author's biography, and unsupported business claims are excluded.

1. **Choose a useful topic.** Favor a topic that meets at least two criteria: new, relevant to many people in the selected audience, practical now, or connected to a clear business benefit. This is an editorial filter, not proof of demand. Confirm the intended audience and the one action they should take.
2. **Create the research packet.** Write what changed or what problem is being solved, why this audience cares, a realistic application, and dated source links for factual claims. Use enough reliable sources to support the actual claims. Keep creator statements, verified facts, and hypotheses distinct.
3. **Set an original angle.** Reuse the studied mechanism, such as result before process or proof before objection. Replace the creator's topic, story, footage, and words with the selected brand's truthful material. A soup demonstration can inform a software demonstration's structure without making the software script about soup.
4. **Build in Radar.** Open Inspiration → US trends. Expand an example and choose Build script. Fill in the new topic, audience, duration, viewer promise, angle, verified facts and source links, and one primary CTA. Generate the script through the existing editor, then edit the result before approval.
5. **Write hook first.** Deliver a clear promise in the opening. Follow with only the context needed, why it matters, practical steps or a demonstration, and a useful payoff. Include mistakes only when they help the specific brief. Close with the primary CTA. Short form should not inherit the document's long-form word target.
6. **Review each claim.** Verify numbers, features, dates, policies, testimonials, and promised outcomes. Do not invent a personal story or treat another creator's success as the brand's proof. Remove filler, long spoken sentences, unclosed loops, and hooks the body does not deliver. Keep missing proof visible instead of filling it with certainty.
7. **Package the result.** Produce accurate title or opening-text options, a cover concept, caption, and relevant keywords when appropriate. Keep annotated research separate from clean narration. Use the existing editor for hooks, spoken copy, shot planning, and caption review.
8. **Learn from published results.** Use owned analytics to compare opening performance, viewing duration or completion, audience responses, and CTA outcomes. Revise the specific weak element. Do not claim a retention improvement from public view counts alone.

### Reusable script brief

- Topic and selected source study:
- Audience and platform:
- Target duration:
- Viewer promise:
- Original angle:
- Verified facts and dated source URLs:
- Proof or demonstration available:
- Primary CTA:
- Three original hook options:
- Ordered beats and payoff:
- Claims requiring review:
- Packaging notes:

A duration target is a writing estimate. Read the finished script aloud and adjust pacing before recording. Keep the research context available to the writer without inserting research annotations into spoken copy.

## Primary sources

Checked 2026-10-15:

- [ScrapeCreators TikTok keyword search](https://docs.scrapecreators.com/v1/tiktok/search/keyword/): US proxy semantics, dates, sorting, pagination.
- [ScrapeCreators Instagram recent popular Reels](https://docs.scrapecreators.com/v1/instagram/reels/trending/openapi.json): seven-day verification and popularity limits.
- [ScrapeCreators Instagram search](https://docs.scrapecreators.com/v2/instagram/reels/search/): Google indexing and date-window limits.
- [TikTok Creative Center Trends](https://ads.tiktok.com/creative/creativeCenter/trends?LanguageId=1): regional trend context. Verify the rendered country and period.
- [TikTok Creative Codes](https://ads.tiktok.com/business/en-US/creative-codes): hook, body, close, and attention devices. This is advertising guidance.
- [TikTok recommendations](https://support.tiktok.com/en/using-tiktok/exploring-videos/how-tiktok-recommends-content): watch behavior and other inputs, without a public scoring formula.
- [Instagram Reels insights](https://about.fb.com/news/2023/04/instagram-reels-trending-audio-and-gifts-updates/): watch time and improving openings.
- [Instagram Trial Reels](https://about.fb.com/news/2024/12/trial-reels-try-content-non-followers-first-see-what-perfoms-best/): testing content with non-followers when available.